XML is notoriously expensive to properly parse in many languages. Basically, the entire world centers around 3 open source implementations (libxml2, expat and Xerces), if you want to get anywhere close to actual compliance. Even with them, you might hit challenges (libxml2 was largely unmaintained recently, yet it is the basis for many bindings in other languages).
The main property of SGML-derived languages is that they make "list" a first class object, and nesting second class (by requiring "end" tags), and have two axes for adding metadata: one being the tag name, another being attributes.
So while it is a suitable DSL for many things (it is also seeing new life in web components definition), we are mostly only talking about XML-lookalike language, and not XML proper. If you go XML proper, you need to throw "cheap" out the window.
Another comment to make here is that you can have an imperative looking DSL that is interpreted as a declarative one: nothing really stops you from saying that
means exactly the same as the XML-alike DSL you've got.
One declarative language looking like an imperative language but really using "equations" which I know about is METAFONT. See eg. https://en.wikipedia.org/wiki/Metafont#Example (the example might not demonstrate it well, but you can reorder all equations and it should produce exactly the same result).
Or... you could just use a programming language that looks good and has great support for embedded domain-specific languages (eDSL), like Haskell, OCaml or Scala.
Or, y'know, use the language you have (JavaScript) properly, eg. add a sum abstraction instead of .reduce((acc, val) => { return acc+val }, 0).
In particular, the problem of "all the calculations are blocked for a single user input" is solved by eg. applicatives or arrows (these are fairly trivial abstract algebraic concepts, but foreign to most programmers), which have syntactic support in the abovementioned languages.
(Of course, avoid the temptation to overcomplicate it with too abstract functional programming concepts.)
If you write an XML DSL:
1. You have to solve the problem of "what parts can I parallelize and evaluate independently" anyway. Except in this case, that problem has been solved a long time ago by functional programming / abstract algebra / category-theoretic concepts.
2. It looks ugly (IMHO).
3. You are inventing an entirely new vocabulary unreadable to fellow programmers.
4. You will very likely run into Greenspun's tenth rule if the domain is non-trivial.
Basically, a node is an object with one entry, whose key is the type and whose value is an array. It's a rather S-expressiony approach. if you really don't like using arrays for all the contents, you could always use more normal values at the leaves:
It has the nice property that you're always guaranteed to see the type before any of the contents, even if object keys get reordered, so you can do streaming decoding without having to buffer arbitrary amounts of JSON. Probably not important when parsing a tax code, but can be useful for big datasets.
While a great article, I actually found this linked post [0] to be even better, in which the author lays out how so much modern tooling for web dev exists simply because XML lost the browser war.
EDIT: obviously, JSON tooling sprang up because JSON became the lingua franca. I meant that it became necessary to address the shortcomings of JSON, which XML had solved.
S-expressions are a cheap dsl too. I use it in my desktop browser runtime that is powered by wasm that I’m developing
As the “HTML”^1 and CSS^2 in fact it works so well I use it also reused it to do the styling for html exports in my markup language designed to fight documentation drift^3.
XML is beloved by tax authorities. The Polish tax authorities really love their e-documents and online filing. Except their XML documents are completely human-unreadable, since the schemas are based on field numbers in paper forms. Even in the brand new National e-Invoicing System, designed from scratch, with no paper forms, most fields have names like ‹P_19N›1‹/P_19N›. You read the XML schema to find out it is a "Marker of lack of delivery of goods or provision of services exempt from tax under Article 43 paragraph 1 of the [VAT] Act, Article 113 paragraphs 1 and 9 of the Act or regulations issued under Article 82 paragraph 3 of the Act or under other provisions" (Google Translated, because of course everything is in Polish). So my invoice is saying "yes [1], I am not [N] exempt from tax under $allThatNonsense [P_19]".
In unrelated news, the main author of the VAT Act is offering tax consulting services, as Registered Tax Advisor #00001.
It's not a DSL. It's a generic lexer and parser. It takes the text and gives you an abstract syntax tree. The actual DSL is your spec, and the syntax you apply.
It's one of many equivalent such parser tools, a particularly verbose one. As such it's best for stuff not written by hand, but it's ok for generated text.
It has some advantages mostly stemming from its ubiquity, so it has a big tool kit. It has a lot of (somewhat redundant) features, making it complex compared to other options, but sometimes one of those features really fits your use case.
It was also about how easy it was to generate great XML.
Because it is complicated and everyone doesn't really agree on how to properly representative an idea or concept, you have to deal with varying output between producers.
I personally love well formed XML, but the std dev is huge.
Things like JSON have a much more tighter std dev.
The best XML I've seen is generated by hashdeep/md5deep. That's how XML should be.
Financial institutions are basically run on XML, but we do a tonne of work with them and my god their "XML" makes you pray and weep for a swift end.
Given that that is had strong schema XSD verification built in, where you can tell in an instant whether or not the document is correct; it’s the right tool for a majority of jobs.
My experience has been the people complaining about it were simply not using automated tools to handle it. It’s be like people complaining that “binaries/assembly are too hard to handle” and never using a disassembler.
XML makes for a pretty good markup language and an ok data interchange format(not a great fit, but the tooling is pretty good). but every single time I have seen it used as a programing language I found it deeply regrettable.
For comparison JSON is a terrible markup language, a pretty good data interchange format, and again, a deeply regrettable programing language. I don't know if anyone has put programing language in straight JSON (I suspect they have shudders) but ansible has quite a few programing structures and is in YAML which is JSON dressed in a config language's clothes.
However as a counter point to my json indictment, it may be possible to make a decent language out of it, look to lisp, it's S-expressions are a sort of a data interchange format(roughly equivalent to json) and it is a pretty good language.
> It evokes memories of SOAP configs and J2EE (it’s fine, even good, if those acronyms don’t mean anything to you).
Heh, a couple of years ago I walked past a cart of free-to-take discards at the uni, full of thousand-page tomes about exciting subjects like SOAP, J2EE and CORBA. I wonder how many of the current students even recognized any of those terms.
It's completely unbelievable that so-called developed countries are struggling with this in 2026.
In Norway, we've had a more or less automated tax system for many years; every year you get a notification that the tax settlement is complete, you log in and check if everything is correct (and edit if desired) and click OK.
I like this post, but I gotta tell you, it just makes me want to dust off and write a bunch of s-expr tools to make that ecosystem equally or more attractive for DSLs.
If I do, the IRS will be the first to know about it! I'll staple an announcement to my 1040. ;-)
The article mentions prolog but doesn't mention you can use constraints to fully express his computation graph.
My prefered library is clpBNR which has powerful constraints over boolean, integers and floats:
If you restrict yourself to the pure subset of prolog, you can even express complicated computation involving conditions or recusions.
However, this means that your graph is now encoded into the prolog code itself, which is harder to manipulate, but still fully manipulable in prolog itself.
But the author talks about xml as an interchange format which is indeed better than prolog code...
It kinda blows my mind that after XML we've managed to make a whole bunch of stuff that's significantly worse for any serious usage.
JSON: No comments, no datatypes, no good system for validation.
YAML: Arcane nonsense like sexagesimal number literals, footguns with anchors, Norway problem, non-string keys, accidental conversion to a number, CODE INJECTION!
I don't know why, but XML's verbosity seems to cause such a visceral aversion in a lot of people that they'd rather write a bunch of boring code to make sure a JSON parses to something sensible, or spend a day scratching their head about why a minor change in YAML caused everything to explode.
Actually my own problem with XML was annoyance that back when I had the thought of doing a complex config format in XML, the idea of modifying it programmatically while retaining comments turned out to be absolutely non-trivial. In comparison with the mess one can make with YAML that's just a trivial thing.
1. standardize on JSON as the internal representation, and
2. write a simple (<1kloc) Python-based compiler that takes human-friendly, Pythonic syntax and transforms it into that JSON, based on operator overloading.
So you would write something like:
from factgraph import Max, Dollar # or just import *
tentative_tax_net_nonrefundable_credits = Max(Dollar(0), total_tentative_tax - total_nonrefundable_credits)
Values like total_nonrefundable_credits would be objects of class Node that "know where they come from", not imperatively-calculated numbers. The __sub__ method (which is Python's way of operator overloading) would return a new node when two nodes are subtracted.
XML (and those prolog and KDL expressions) have one big advantage over JSON: The type (in XML the tag name) comes before the rest of the object. In JSON it's usually a type field. That means in JSON it could come at any point and thus you have to load the whole sub-structure as a dynamic hash map before you can evaluate the type field and instantiate the correct type in your programming language. With XML using a SaX parser you are guaranteed to get the type first and thus can immediately instantiate the correct type and load the properties into that, skipping any dynamic hash map. Depending on your application this can mean a big performance difference.
There is a middle ground between using XML and imperative code for representing tax forms. Robert Sesek’s ustaxlib [0] uses JavaScript to encode the forms in a way that is reasonably statically analyzable. See the visualizer [1]. My approach uses XML to represent the forms with an embedded DSL to represent most expressions tersely. See for example Form 8960 in ustaxlib [2] and my TaxStuff program [3]. The main thing that the XML format from the article has going for it is that it is easy to write a parser for. But it is a bit verbose for my taste.
I like how this article lists various alternatives. Like I was thinking "well, JSON is more compact", and they covered JSON. And then "well, s-expressions supports nesting too", and then they covered s-expressions as well. The best documentation always include the things that weren't done.
265 comments
The main property of SGML-derived languages is that they make "list" a first class object, and nesting second class (by requiring "end" tags), and have two axes for adding metadata: one being the tag name, another being attributes.
So while it is a suitable DSL for many things (it is also seeing new life in web components definition), we are mostly only talking about XML-lookalike language, and not XML proper. If you go XML proper, you need to throw "cheap" out the window.
Another comment to make here is that you can have an imperative looking DSL that is interpreted as a declarative one: nothing really stops you from saying that
means exactly the same as the XML-alike DSL you've got.One declarative language looking like an imperative language but really using "equations" which I know about is METAFONT. See eg. https://en.wikipedia.org/wiki/Metafont#Example (the example might not demonstrate it well, but you can reorder all equations and it should produce exactly the same result).
Or, y'know, use the language you have (JavaScript) properly, eg. add a
sumabstraction instead of.reduce((acc, val) => { return acc+val }, 0).In particular, the problem of "all the calculations are blocked for a single user input" is solved by eg. applicatives or arrows (these are fairly trivial abstract algebraic concepts, but foreign to most programmers), which have syntactic support in the abovementioned languages.
(Of course, avoid the temptation to overcomplicate it with too abstract functional programming concepts.)
If you write an XML DSL:
1. You have to solve the problem of "what parts can I parallelize and evaluate independently" anyway. Except in this case, that problem has been solved a long time ago by functional programming / abstract algebra / category-theoretic concepts.
2. It looks ugly (IMHO).
3. You are inventing an entirely new vocabulary unreadable to fellow programmers.
4. You will very likely run into Greenspun's tenth rule if the domain is non-trivial.
EDIT: obviously, JSON tooling sprang up because JSON became the lingua franca. I meant that it became necessary to address the shortcomings of JSON, which XML had solved.
0: https://marcosmagueta.com/blog/the-lost-art-of-xml/
1. https://gitlab.com/canvasui/canvasui-engine/-/blame/main/exa...
2. https://gitlab.com/canvasui/canvasui-engine/-/blob/main/exam...
3. https://gitlab.com/sablelang/libcuidoc
In unrelated news, the main author of the VAT Act is offering tax consulting services, as Registered Tax Advisor #00001.
It's one of many equivalent such parser tools, a particularly verbose one. As such it's best for stuff not written by hand, but it's ok for generated text.
It has some advantages mostly stemming from its ubiquity, so it has a big tool kit. It has a lot of (somewhat redundant) features, making it complex compared to other options, but sometimes one of those features really fits your use case.
It was also about how easy it was to generate great XML.
Because it is complicated and everyone doesn't really agree on how to properly representative an idea or concept, you have to deal with varying output between producers.
I personally love well formed XML, but the std dev is huge.
Things like JSON have a much more tighter std dev.
The best XML I've seen is generated by hashdeep/md5deep. That's how XML should be.
Financial institutions are basically run on XML, but we do a tonne of work with them and my god their "XML" makes you pray and weep for a swift end.
My experience has been the people complaining about it were simply not using automated tools to handle it. It’s be like people complaining that “binaries/assembly are too hard to handle” and never using a disassembler.
For comparison JSON is a terrible markup language, a pretty good data interchange format, and again, a deeply regrettable programing language. I don't know if anyone has put programing language in straight JSON (I suspect they have shudders) but ansible has quite a few programing structures and is in YAML which is JSON dressed in a config language's clothes.
However as a counter point to my json indictment, it may be possible to make a decent language out of it, look to lisp, it's S-expressions are a sort of a data interchange format(roughly equivalent to json) and it is a pretty good language.
> It evokes memories of SOAP configs and J2EE (it’s fine, even good, if those acronyms don’t mean anything to you).
Heh, a couple of years ago I walked past a cart of free-to-take discards at the uni, full of thousand-page tomes about exciting subjects like SOAP, J2EE and CORBA. I wonder how many of the current students even recognized any of those terms.
In Norway, we've had a more or less automated tax system for many years; every year you get a notification that the tax settlement is complete, you log in and check if everything is correct (and edit if desired) and click OK.
It shouldn't be more difficult than this.
If I do, the IRS will be the first to know about it! I'll staple an announcement to my 1040. ;-)
But the author talks about xml as an interchange format which is indeed better than prolog code...
JSON: No comments, no datatypes, no good system for validation.
YAML: Arcane nonsense like sexagesimal number literals, footguns with anchors, Norway problem, non-string keys, accidental conversion to a number, CODE INJECTION!
I don't know why, but XML's verbosity seems to cause such a visceral aversion in a lot of people that they'd rather write a bunch of boring code to make sure a JSON parses to something sensible, or spend a day scratching their head about why a minor change in YAML caused everything to explode.
Actually my own problem with XML was annoyance that back when I had the thought of doing a complex config format in XML, the idea of modifying it programmatically while retaining comments turned out to be absolutely non-trivial. In comparison with the mess one can make with YAML that's just a trivial thing.
1. standardize on JSON as the internal representation, and
2. write a simple (<1kloc) Python-based compiler that takes human-friendly, Pythonic syntax and transforms it into that JSON, based on operator overloading.
So you would write something like:
and then in class Node (in the compiler): Values like total_nonrefundable_credits would be objects of class Node that "know where they come from", not imperatively-calculated numbers. The __sub__ method (which is Python's way of operator overloading) would return a new node when two nodes are subtracted.[0]: https://github.com/rsesek/ustaxlib
[1]: https://github.com/rsesek/ustaxviewer
[2]: https://github.com/rsesek/ustaxlib/blob/master/src/fed2019/F...
[3]: https://github.com/AustinWise/TaxStuff/blob/master/TaxStuff/...